C++ SCHX ยท rev 4
Reasoning ยท v0.24.0

Deliberation and task executive

How ABRIA creates an explicit task model, complementary cognitive roles and a completion contract.

SCHX / FEATURE LAYER

Source map

  • python/abria_server/intelligence.py
  • python/abria_server/abr_deliberation.py
SCHX / FEATURE LAYER

Task model

IntelligenceExecutive derives topics, intents, constraints, goals, procedures, verification questions and anti-patterns from the current request plus learned patterns/tool plan.

SCHX / FEATURE LAYER

Complementary roles

Depending on the task it can add roles such as diagnostician, skeptic, evidence-scout, tool-router, constraint-auditor and implementation-engineer, while deduplicating near-identical roles.

SCHX / FEATURE LAYER

Parallelism budget

Complexity, uncertainty and role count determine a bounded parallelism budget rather than spawning arbitrary numbers of workers.

SCHX / FEATURE LAYER

Completion contract

The task executive builds explicit success criteria so the final answer can be checked against the user's requirements before it is returned.